Red Flags in Development Agency Proposals: What to Check Before You Sign

⚡The short answer: the flags that matter most

A proposal is a prediction of a relationship. Agencies behave during the sale the way they will behave during the build — only nicer. So when a proposal hides information, the build will hide it too, at build rates. The three flags that predict the most damage are a single-point estimate with no discovery, an anonymous team, and IP language that does not assign everything to you on payment. Almost everything else is a variation on those three.

The table below is the checklist version. The rest of the article explains each flag, why it happens, and what to ask in response. None of these are automatic disqualifiers on their own — a small fixed-scope job legitimately skips a formal discovery phase, for example — but each one is information the agency chose not to give you, and the pattern matters more than any single item.

See how our services are scoped and priced→

📐Flags one to three: the estimate itself

The estimate section of a proposal is where the agency shows you whether it understands your project or is selling to it. Three patterns should slow you down.

First, no discovery phase. A professional build starts with a paid, time-boxed scoping phase — typically one to four weeks — that converts your idea into a buildable backlog with real estimates. A proposal that jumps straight to a build total is telling you the number was produced from a sales call, not from analysis. The discovery cost did not disappear; it moved into the build, where it will be spent as rework at full rates. Skipping discovery is sometimes legitimate for a genuinely small, well-defined job, but the proposal should say so explicitly.

Second, a single confident total. Real estimation has variance, and honest agencies show it: a range, or a total with stated assumptions. A lone number like $87,500 for a six-month build signals false precision. The variance still exists — you just cannot see it, which means it will arrive later as change orders, and change orders are priced by a vendor who knows switching costs are now high.

Third, an estimate that cannot decompose. Ask for hours by role and phase. An agency that cannot or will not break its total into engineering, design, QA and management hours is asking you to evaluate a number you cannot verify or negotiate. The refusal itself is the flag — even if the total is fair.

👥Flags four to seven: who is actually doing the work

You are not buying a company; you are buying specific people for specific months. Proposals that blur this are protecting the agency, not you.

A vague team page — "our senior architects" with no names, or a wall of stock photos — usually means one of two things: staffing is not settled, or the people on the sales call are not the people who will build your product. The counter is simple: ask for the named team, their roles, their allocation percentages, and the right to interview the technical lead. Agencies with stable senior teams agree to this readily; agencies running a bench or a resale operation stall.

The next flag is "100 percent local" staffing that is actually offshore. There is nothing wrong with offshore or blended teams — we run one ourselves, with offices in Edmonton and Chandigarh, and the model works well when it is disclosed and managed. The red flag is the lie, not the geography. An agency claiming a fully local team while quietly subcontracting overseas is billing you onshore rates for offshore delivery and, more importantly, has already shown you how it handles uncomfortable truths. Ask where each named person physically sits and check it on a video call.

Two smaller flags round out this group. No QA line item means testing has no budget owner, and under deadline pressure unbudgeted work is what gets cut — you receive code that was never systematically tested and discover it in production. And missing maintenance terms mean month thirteen is unpriced, which hands the agency open-ended leverage exactly when switching is most painful. Honest proposals name a maintenance model and a rate before you sign.

📄Flags eight to eleven: the contract mechanics

The final group lives in the terms rather than the pitch, and it is where sophisticated buyers focus. These four are the ones that produce the worst surprises.

IP assignment gaps are the most consequential. The contract should assign all work product — code, design files, documentation, infrastructure definitions — to you, in writing, upon payment, with no retained licence that lets the agency reuse your product as a template for a competitor. Watch for: assignment only on final payment of the last invoice (leverage during disputes), exclusions for "pre-existing tools" defined so broadly they swallow the codebase, and silence on design and infrastructure assets. Anything less than full assignment is renting, not buying.

An unrealistic timeline is a sales flag disguised as optimism. If three agencies quote five months and one quotes ten weeks, the ten-week proposal has not found an efficiency the others missed — it has deferred the honest conversation to a missed deadline, when your sunk cost makes you compliant. Ask what the timeline assumes about your own feedback turnaround, and what has made similar projects slip.

The last two travel together: no change process and a proprietary lock-in stack. Every real project changes scope; a proposal without a written process for pricing and approving changes is reserving the right to improvise billing later. And a proposal built on a niche framework or a vendor-controlled platform nobody else can maintain is a switching-cost strategy. A mainstream stack — mainstream frameworks, a standard database, managed cloud — is cheaper to build, cheaper to maintain, and, tellingly, the choice of agencies that expect to keep your business by being good rather than by being sticky.

✅What a good proposal actually contains

It is easier to spot the flags when you know the shape of the real thing. A professional proposal is longer than a sales deck and shorter than a contract, and it is specific in ways that are mildly uncomfortable for the agency — because specificity is what accountability looks like before signature.

Notice what the good shape has in common across every row: it transfers information to you before you have leverage over nothing. Named people, decomposed hours, written exclusions, priced maintenance. An agency that hands you this document has deliberately given up the informational advantages that bad agencies monetize later. That is not naivety; it is a firm confident enough in its delivery to compete on it.

The table also works in reverse as a request list. If a proposal arrives missing two or three of these rows, you do not have to reject it — you can ask for the rows. The quality of the response to that request is itself the evaluation. A firm that fills the gaps quickly and precisely has them ready; a firm that cannot produce them was never going to.

Read what our proposals include as standard→

⚖️How to compare two proposals honestly

Three proposals for the same project routinely span a two-times price range, and the middle one is not automatically right. Before comparing totals, normalize. Force every proposal onto the same grid: same team mix assumption, same phase split, same list of what is included and excluded. The classic exclusions — hosting, third-party licences, content entry, data migration, post-launch support — are where a cheap proposal quietly becomes expensive. An excluded line you need is a discount you did not get.

Then compare the blended rate, not the headline. Divide each total by its total hours. A $150,000 proposal with senior-heavy staffing and full QA frequently represents better value than a $110,000 proposal that is mostly junior hours with no testing budget — because the second one delivers the missing phases anyway, as rework, at the worst possible time.

Weight the things proposals cannot show. Responsiveness during the sale predicts responsiveness during the build. The quality of the questions the agency asked you predicts the quality of the discovery. And the willingness to say "this part of your scope is a bad idea" predicts whether you are buying a partner or a meter.

Finally, run the downside scenario explicitly. Ask each agency: "at 80 percent of the budget, if the remaining work is larger than estimated, what happens?" The answer — re-plan with you, or bill the overrun — tells you which proposal was real. This one question filters more effectively than any scoring matrix, because it forces the agency to describe its own failure mode before you have signed.

❓The ten questions to ask before signing

Everything above compresses into a short list of questions that any honest agency will answer directly and any problematic one will hedge on. Ask them in writing, and treat hedged written answers as answers.

One discipline makes the list far more powerful: ask every shortlisted agency the same questions and compare the answers side by side. Differences that were invisible in the proposals become obvious in the responses — one agency names people, another describes roles; one quotes a maintenance rate, another says "we can discuss that closer to launch." You are not just collecting answers; you are generating a controlled comparison of how each firm behaves under the same mild pressure.

Start a scoped conversation about your project→